How to give your dot responsibilities

Your dot is more powerful than you’re giving it credit for. It’s equipped with its own browser, Codex, the plugins you connect to it, and the ability to schedule itself reminders. Putting it this way, it’s not all that different from anyone else who uses a computer.

But if your experience is like that of the users I’ve talked to, you’ve probably experienced some version of these two things:

1. Proactive moments of “small delight.” Your dot gives you a heads up that a package arrived. Nice to know, though it would’ve been at your doorstep anyway.

2. Prompted, momentous efforts. You call your dot at the airport and say “I want you to fight United tooth and nail to get this right until my entire flight has been comped” (possibly personal experience).

I’d put some of my favorite things my dot does in a third category: ongoing responsibilities.

I travel a lot for work. Every month or so I’d ask ChatGPT to find the food I bought on work trips, my L&D subscriptions, all those little expenses, and send them to my work email. Then I’d ask my work ChatGPT to expense them on Navan.

Now my work dot sends trip invites to my personal calendar so my personal dot knows when I’m traveling. It keeps an eye on my Finances and emails the expenses to my work dot every week, with receipts when available. My work dot checks that they’re in policy and expenses them on Navan. I hear about successful reimbursements and anything that needs me. The rest is out of mind.

Here’s what I mean by a responsibility, and how to hand one off to your dot so you don’t have to keep thinking about it.

Responsibilities vs Tasks

The progression from “AI that completes tasks” to “AI that has responsibilities” has been beaten to a pulp, so I will keep this brief.

Tasks are prompted by the user. And through a combination of tools, plugins, and memory, agents complete them. “Turn the button red.” “Update the deck and send it to Jamie.” “Book me a dinner reservation at the Olive Garden b/t 6 and 7pm for two next Tuesday.” Input from user, output from agent, task complete ✅.

If I still have to tell my dot every time there’s a launch, then I’m still the one responsible for keeping an eye on launches. We all have responsibilities. “Take the garbage out every Wednesday before 7AM.” “Get all alpha participants early access before the release.” “Make sure my work expenses get submitted every week.”

Some responsibilities are short-term in nature. Consider them “projects”. When I asked my dot to find me movers, I told it, “I want you to own that responsibility and don’t stop until I’ve secured movers at a good rate.” Same goes for finding a photographer for an event or a contractor for a home project. Once you’ve found the right person and secured the date, you’re done.

Other responsibilities are indefinite. My work expenses are one of those. Keeping an eye on launches so I have a quote tweet ready to go is another. So is checking my calendar for work trips and making sure my personal dot has the dates.

But here’s the thing: your dot can handle responsibilities (it already does). Where things get muddy, however, is when the specificity and clarity on those responsibilities is lost, and overlapping instructions create a nasty web of tasks for your dot to carry out.

If you’re finding that your dot is sending you messages that aren’t helpful more often than not, this is for you.

Your dot needs clear expectations

It’s hard to know what good looks like when you don’t know what’s expected of you. Same goes for your dot. It just wants to be helpful!

So, I plead with you to ask your dot the following:

What are you keeping an eye on for me right now? Give me a short list of the responsibilities you’re tracking. For each one, tell me what I asked you to handle, when you check, and when I should expect to hear from you. Flag anything we talked about but never set up.

If your dot is like mine, I expect it will return with a laundry list of inter-connecting tasks, most of which you don’t remember asking for and most of which don’t have clear “good” outcomes.

Go through it together, and tell your dot what to keep, what overlaps, and what you want it to stop.

I wanted to start from scratch. If you do too, here’s what you ask next:

Let’s start over. Stop the responsibilities you’re currently tracking for me and turn off the reminders and checks that go with them. Keep your memory of me and our conversations. Let me know when we’re ready to start fresh, and flag anything you couldn’t turn off.

How to give your dot responsibilities

I wanted to create a system for responsibility delegation that actually works. I think the right recipe for this is pretty simple.

1. A shared document with a table of responsibilities. Mine is a Page in Space where I can add them and my dot can add them as I instruct.

2. A prompt that tells my dot to use that Page, set up the work, and keep me in the loop when something needs me.

First, ask your dot to create the Page:

Create a private Page in Space called Responsibilities, with a table of the responsibilities you’re tracking and a short Current work section. If we already have that Page, use it. Send me the link.

Put the Page link in these instructions and send them to your dot:

Use our Responsibilities Page at
[paste the Page link here].

Keep these instructions on the Page so we can refer
back to them the next time I hand something off.

Read it before you act on a responsibility or change
one. Keep one row per responsibility, with these
columns:

- Responsibility — what I’m handing off and the
  outcome I want.
- Scope and actions — where to find the information,
  what you can handle on my behalf, and what needs
  my approval.
- Trigger and timing — what to watch for or when to
  check. Include the time zone if there’s a
  schedule.
- Done when — what a good result looks like and how
  you’ll check it. For recurring work, what counts
  as done this time?
- Communication — when to message me, where to reach
  me, and when to stay quiet.
- Duration and review — when to stop, or how long to
  keep doing this before we look at it again.
- Status and execution — is it proposed, active,
  blocked, paused, or done? Link the actual check or
  schedule and include the last result you verified.

When I hand something off, check for overlap first.
If it’s a one-off, just do the task. If you need
something from me to get going, ask. Set up the
agreed check and verify it before marking the row
active. If you can’t set it up, tell me what’s
missing.

Under Current work, show what you’re doing now, what
happens next, and anything that needs me. Check the
result before you call it done. When I pause or
retire a responsibility, stop the related checks and
update the row.

If you see something else you could take off my
plate, suggest it first. If we agree on a regular
review, give that its own row and flag anything
that’s stopped working or overlaps.

Here’s what you can expect going forward:

• you and your dot are now on the same page (literally) regarding the responsibilities you want it to carry out

• you can see what your dot is working on, and it can suggest responsibilities you might want to hand off next

What responsibilities should I give my dot?

Start with the accumulated context you’ve built up already

Based on our experience together, what are the top 5 responsibilities you could take on for me to make my life easier?

If none of those are appealing, give these a try:

1. Every Sunday night, check both our calendars against the school emails. Tell us which pickups or drop-offs we still need to figure out and if the bus schedule changed.

2. Keep the recipes I send you during the week. Pick a few for next week and put the ingredients in an Instacart cart for me to review. Then remind me what I was going to cook.

3. Whenever I have a flight coming up, keep an eye on the plane coming in for it. If that plane’s running late, send me my same-day options while I still have time to switch.

4. Keep an eye on the free trials I sign up for. A few days before one starts charging me, tell me what it is and what it’ll cost. If I say cancel, handle it for me.

5. Keep track of emails where I’m waiting on someone. If it’s been two business days, check whether they got back to me somewhere else. If I’m still waiting, draft a follow-up in the thread.

6. When a package needs a signature, check my calendar. If it looks like I’ll be out, give me a heads up while I still have time to change the delivery.

7. Keep an eye on home games for my team. When tickets go on sale, check my calendar and send me the link for the games I can actually make.

8. When someone emails me to find a time, read the thread to see who needs to be there, check my calendar, and handle the back-and-forth. If you can’t tell who should be on the invite, ask me.

9. Keep a list of the parts I’m looking for and what they usually cost. Keep an eye on the shops I buy from. If something on my list comes back in stock or goes on sale, send me the price and the link.

10. Whenever I send you footage for a training video, edit it the way we did the last one. Send me a cut to review and use my feedback the next time I send footage.

If you try this, let me know what you gave your dot. I’m curious what you used to have to keep an eye on that you don’t even think about anymore.